International Journal on Recent and Innovation Trends in Computing and Communication

Volume: 2 Issue: 5

ISSN: 2321-8169
1130 1135

_______________________________________________________________________________________________

VLSI Architecture Design for Bi-orthogonal Wavelet Filters using Algebraic

Integer Encoding
P.Vijayalakshmi1, M.Vidya2
(Department of Electronics and Communication Engineering, Final year ME (Applied Electronics), Sri Venkateswara College of
Engineering, Sriperumbudur) 1
(Department of Electronics and Communication Engineering, Assistant Professor, Sri Venkateswara College of Engineering,
Sriperumbudur) 2
Abstract--As the world advances with technology and research, images are widely used in many fields such as biometrics, remote sensing,
reconstruction etc. This growth in image processing applications, demands majorly for low power consumption, low cost and small chip area.
Thus the advances in the VLSI system provides powerful tool for complicated imaging systems. A flexible architecture presents for algebraic
integer based encoding of bi-orthogonal wavelet filters. A single final reconstruction step of algebraic integer encoding provides filtered and
down sampled image outputs resulting in low levels of quantisation noise. Filter coefficients of the bi-orthogonal wavelet filters are quantized
before implementation. In this architecture, all multiplications are performed using less shifts and additions. MSE and PSNR values of algebraic
integer encoding provides improvement when compared to the fixed point representation. This architecture also provides better performance in
terms of reducing chip area when compared to the Fixed point Technique.
Keywords: Biorthogonal wavelet Transform, Fixed Point Approximation, Algebraic Integer Encoding

__________________________________________________*****_________________________________________________

I.

Introduction

Recent advances in medical imaging and

telecommunication systems require high speed, resolution
and real-time memory optimization with maximum
hardware utilization. The 2D Discrete Wavelet Transform
(DWT) is widely used method for these medical imaging
systems because of its reconstruction property. Discrete
Wavelet Transform
can decompose the signals into
different sub bands with both time and frequency
information and facilitate to arrive at maximum compression
ratio. Wavelet architecture, in general, reduces the area
requirements and enhances the speed of communication by
breaking up the image into the blocks.
VLSI has been around for a long time, there is
nothing new about it, but as a side effect of advances in the
world of computers, there has been a dramatic proliferation
of tools that can be used to design VLSI circuits. A wavelet
is a mathematical function used to divide a given function or
continuous-time signal into different scale components.
Usually one can assign a frequency range to each scale
component. Each scale component can be studied with a
resolution that matches its scale.
A wavelet transform is the representation of a
function by wavelets. The wavelets are generally scaled and
translated copies are known as daughter wavelets of fastdecaying oscillating waveform known as the mother
wavelet. Wavelet transforms have advantages when
compared to current Fourier transforms for representing
functions that have breaks and sharp peaks, and for
accurately analysing and recreating finite, non-periodic and
non-stationary signals. Generally, an estimate to Discrete
Wavelet Transform is used for data compression and the
CWT is used for signal analysis. Thus, DWT is commonly
used in engineering and computer science, and CWT in

scientific research. Wavelet transforms are now being

applicable for many number of applications, often replaced
the current Fourier Transform.
Many areas of physics have seen this paradigm
swing, including molecular dynamics, ab initio designs,
astrophysics, density-matrix localisation, seismology, optics,
commotion and quantum mechanics. This change has also
happened in image processing, blood-pressure, heart-rate,
brain rhythms, DNA analysis, protein analysis, climatology,
general processing of signal, speech recognition, computer
graphics and multifractal analysis. In computer vision and
image processing, the idea of scale-space representation and
Gaussian derivative operators is regarded as a canonical
multi-scale representation.

II.

Related Work

Many architectures have been proposed to perform

DWT, but few address the precision of the coefficients
necessary to ensure perfect reconstruction. The goal of this
work is to determine the precision of the filter coefficients
(for an orthogonal wavelet) needed to compute the 2-D
DWT without introducing round-off error via the filter. It
needs at least 14 bits for 1 octave of decomposition if both
forward and inverse DWT are fixed-point operation, while
13 bits are enough to correlate with the same result if only
the transform DWT is fixed-point[2]. The implementation of
2D DWT using fixed point representation are shown and the
results shown that this can be performed without any loss [3].
VLSI Architectures for the 4-Tap and 6-Tap 2-D
Daubechies Wavelet Filters Using Algebraic Integers,
offered an algebraic integer (AI) based multi-encoding of
Daubechies-4 and -6 2-D wavelet filters having error-free
integer-based computation. It also guarantees a noise-free
1130

IJRITCC | May 2014, Available @ http://www.ijritcc.org

_______________________________________________________________________________________

International Journal on Recent and Innovation Trends in Computing and Communication

Volume: 2 Issue: 5

ISSN: 2321-8169
1130 1135

_______________________________________________________________________________________________
computation throughput the multi-level multi-rate 2-D
filtering operation. Comparisons are provided between
Daubechies-4 and -6 in terms of several parameters [1].

III.

PROPOSED WORK

Fig.1 shows the block diagram of the proposed

system. Initially the image is given to the bi-orthogonal
wavelet filter, where the image is decomposed based on sub
band coding. The coefficients which we got from the biorthogonal wavelet filter are quantized using fixed point
approximation. Finally the image is reconstructed.

function, there are two scaling functions

2D BIORTHOGONAL
WAVELET
FILTER

FIXED
POINT
REPRESENT
ATION

RECON
STRUC
TED
IMAGE

that may

generate different multiresolution study, and accordingly

two different wavelet functions , is used in the analysis
and is used in the synthesis. In addition, the scaling
functions , and the wavelet functions , are related by
duality in the following equations (1) and (2)

as soon as

j ,k

( x)~ j,k ( x)dx 0

(1)

j j or k and even.

INPUT
IMAGE

, ~

o ,k

( x) 0,k ( x)dx 0

(2)

as soon as k k
There are two sequences,

g n and hn to act as

Fig.1 Block diagram of proposed system

decomposition sequences and two sequences to act as

SUB-BAND CODING

reconstruction sequences. If c1n is a data set, it should be

decomposed as which is shown in the equation (3) and (4)

Generally image is decomposed based on sub-band

coding. This sub-band coding is the general form of biorthogonal wavelet transform. If the scaling and wavelet
functions are separable, the summation can be decomposed
into two stages. First step is along the x-axis and then along
the y-axis. For each axis, we can apply wavelet transform to
increase the speed. The two dimensional signal (usually
image) is divided into four bands: LL (left-top), HL (righttop), LH (left bottom) and HH (right-bottom). The HL band
indicated the variation along the x-axis while the LH band
shows the y-axis variation. The power is more compact in
the LL band. Fig.2 shows the single level decomposition

Fig.2 First level of decomposition

IV.

BI-ORTHOGONAL WAVELET
TRANSFORM

Already known bases that span a space do not have

to be orthogonal. In order to gain greater tractability in the
construction of wavelet bases, the orthogonality condition is
unperturbed allowing semi-orthogonal, bi-orthogonal or
non-orthogonal wavelet bases. Bi-orthogonal Wavelets are
families of compactly supported symmetric wavelets. The
regularity of the filter coefficients is often desirable since it
results in linear phase of the transfer function. In the biorthogonal case, rather than having one scaling and wavelet

cn0 h2 nk c1k

(3)

d n0 g 2 nk c1k

(4)

For reconstruction it is shown as

~
cl1 h2 nl cn0 g~2 nl d n0

(5)

If we want perfect reconstruction, so decomposing

and then reconstructing shouldnt change anything. This
imposes some conditions shown in equation (6) & (7)

~
g n (1) n1 hn

(6)

g~n (1) n1 hn l

(7)

The separation of analysis and synthesis is such

that the useful properties for analysis (e.g., oscillations, zero
moments) can be concentrated on the

function. The

interesting property for synthesis (regularity) which is

assigned to the function has proven to be very useful.
The dual scaling and wavelet functions have the following
properties:

They are zero outside of a segment.

The calculation algorithms are maintained, and
thus very simple.
1131

IJRITCC | May 2014, Available @ http://www.ijritcc.org

_______________________________________________________________________________________

International Journal on Recent and Innovation Trends in Computing and Communication

Volume: 2 Issue: 5

ISSN: 2321-8169
1130 1135

_______________________________________________________________________________________________
The associated filters are symmetrical.
FINAL RECONSTRUCTION STEP

The functions used in the calculations are

easier to build numerically than those used in
the Daubechies wavelets.

FIXED POINT APPROXIMATION

The designs of a wavelet transform processor, to
treat the data and filter coefficient values in fixed-point
values. Fixed point has the advantages of being easier to
implement, requires less silicon area, and makes
multiplications faster to perform. Floating point allows a
greater range of numbers, though floating point numbers
require 32 or 64 bits. Fixed-point numbers can be 8 to 32
bits (or more), possibly saving space in the multiplier.

V.

ALGEBRAIC INTEGER
INTERPRETATION

The motivation for algebraic integers to represent

bi-orthogonal filter coefficients comes from a recently
introduced algebraic integer based encoding scheme that
allows low complexity error free computation. Algebraic
integers are defined by real numbers that are root of monic
3
polynomials with integer coefficients such as 2, 7 etc.
Many research works have been performed using singledimensional and multi-dimensional algebraic integer
encoding of the transform matrix. This technique increases
the sparseness of the encoding matrix, as well as, by making
an appropriate choice of the variables we can reduce the
dynamic range of the transformation coefficients
substantially, which leads to more efficient hardware
implementation.
The whole point of algebraic integer quantization is
not just the reduction of the number of arithmetic operations,
but the lowering of the dynamic range of the computations
which is much more important. This AI encoding is used
which has the advantage of mapping the required irrational
coefficients into arrays of integers. By manipulating these
polynomial representations of the coefficients, instead of
approximate representations of the coefficients themselves,
we have eliminated any errors in the calculations until the
final reconstruction step. Both forward and inverse mapping
can be performed using the same polynomial expansion. In
this reconstruction step appears only once without any
intermediate reconstruction steps.

For the computation of a 2-D DWT or IDWT, we

need to recover the integer part of the result and the most
significant bit of the fractional part, to allow correct
rounding. To produce output results in a conventional
number system, we make a final substitution of the
representation of z in the polynomial representation of the
output data. Since z is irrational, we will not find an errorfree finite representation within a conventional weighted
system (e.g. binary); therefore, we have to use an
approximation, and the error relating to that approximation
dictates the quality of the output result.

VI.

ARCHITECTURE DESIGN

In this paper it is
proposed the improved
architecture in performance by the way of using Algebraic
integer Encoding.

Fig.3 Architecture design of 2D- Discrete Wavelet

Transform
The Fig.3 shows that the architecture of 2D
Discrete Wavelet Transform. Where the input x(n) is given
as image pixels which is stored in the Registers. The image
coefficients are given as input to the low pass and high pass
filters. Then applying Wavelet Transform by passing the
coefficients into the low pass and high pass filter. Then
convolving these coefficients by using multiplier and adder.
The output DWT data c1 is given as input for
applying encoding process. The Fig 4 shows the encoding
process. This architecture design is done by using Xilinx
14.2 software. The encoded data is again applied for IDWT
and then decoding is done.

ADVANTAGES

No floating point calculations,

Absolutely error-free computations (Infinite
precision until final reconstruction step),
Parallel architecture for multiplication which is
very suitable for VLSI implementation,
Less arithmetic calculation which results simplicity
in hardware implementation and finally
The hardware requirements are less in case of AI
compared to classical binary approach.

Fig.4 Algebraic Integer Encoding Process

VII.

RESULTS

1132
IJRITCC | May 2014, Available @ http://www.ijritcc.org

_______________________________________________________________________________________

International Journal on Recent and Innovation Trends in Computing and Communication

Volume: 2 Issue: 5

ISSN: 2321-8169
1130 1135

_______________________________________________________________________________________________
The comparison between the fixed point and
algebraic integer was done by using MATLAB software.
Input image is applied for bi-orthogonal wavelet transform
and then algebraic integer coding is applied. The image is
reconstructed and PSNR, MSE values are noted for various
images. The values shown that there is increase in
performance using algebraic integer encoding.
Architecture is designed using Xilinx 14.2. Initially
image is given as a input and from that input image red, blue
and green components are separated. Each component is
saved as a text. Then these components are given to the
input of the bi-orthogonal wavelet filter. After doing
filtering and encoding operation the output values are
decoded and the image is reconstructed.

Fig.8 Approximation Details

The Table.1 shows that the comparison between fixed point
approximation and Algebraic Integer encoding.
FIXED POINT

ALGEBRAIC
INTEGER

Fig.5 Input Image

Db4

Bior 7/3

Db4

PSNR

15.552 dB

310.47
dB

248.51dB

MSE

1.82e+003

4.66e027

9.23e021

Bior
7/3
311
dB
5.16e027

Table.1 Fixed Point Vs Algebraic Integer

The results shown that Algebraic integer provides
better performance when compared to the Fixed Point
approximation. The Third stage of this project is to design
an Architecture using Xilinx.
Fig. 6 Reconstructed Image
The Fig.5 and 6 shows that the input image is taken
for decomposition and reconstruction using fixed point
approximation. Then PSNR and MSE values are found. The
Fig.7 and 8 shows that input image taken for Algebraic
Integer Encoding.

The main focus to design an architecture for many

image processing applications are the major demands for
power consumption and no of components usage. This
architecture provides better results when compared to the
fixed point approximation.

ARCHITECTURE DESIGN
Designers of digital systems are inevitably faced
with the task of testing their designs. To verify that a design
operates correctly generally simulation is needed, which is
the process of testing design by applying inputs to a circuit
and observing its behaviour. The output of a simulation is a
set of waveforms that show how a circuit behaves based on
a given sequence of inputs.

Fig.7 Input Image

Initially input image is taken and pixel points are

extracted and saved as a text. Then apply DWT process by
writing VHDL program. Then encoding the 8 bit data by
using Algebraic Integer Encoding. Then apply inverse
DWT, decode and finally resize the image. Then the
1133

IJRITCC | May 2014, Available @ http://www.ijritcc.org

_______________________________________________________________________________________

International Journal on Recent and Innovation Trends in Computing and Communication

Volume: 2 Issue: 5

ISSN: 2321-8169
1130 1135

_______________________________________________________________________________________________
reconstructed image is compared with the original image
and PSNR, MSE values are calculated.

Table.2 Comparison of Fixed Point Vs Algebraic Integer

Architecture

The Fig.9 is the input image taken for image

compression. Then from the image Red, Green, Blue
components are extracted. Then these components are saved
as a text file.

The Table.2 shows that decrease in the no of

components used when compared to the proposed system.
This in terms we can say that area is reduced.

Fig.12 Reconstructed Image

Fig.9 Input Image
Then VHDL code is simulated, the text file is given
as a input and the code is simulated using Xilinx Isim
Simulator. The output is stored in the output folder. The
RTL schematic is shown in Fig.10. The simulated timing
diagram for this architecture is shown in Fig.11. After
simulating by entering any key in the matlab command
window, the image is reconstructed and the image is shown
in Fig.12. The RGB components also reconstructed and
several parameters such as compression ratio, PSNR and
MSE values are calculated.

VIII.

A VLSI architecture for computing the AI based

forward and inverse 2-D Discrete Wavelet Transform has
been presented in this paper. The introduced design employs
AI based arithmetic which is error free and free of
multiplications. In order to assess the effectiveness of the
proposed scheme, the design is simulated using Xilinx and
MATLAB. The architecture has been captured by means of
the VHDL language and simulated on data taken from
random images. This architecture is better than the existing
architecture in terms of multipliers and adders which results
in small chip area. Based on the view of high image quality
on reconstructed image, this proposed architecture
outperforms the existing technique. Experimental results
expose the performance of our proposed Algebraic Integer
Encoding technique that accomplishes high PSNR value and
reduced no of components used as compared to the existing
fixed point technique. The same scheme can be extended to
3-D DWT.

IX.
Fig.10 RTL Schematic

CONCLUSION

ACKNOWLEDGEMENT

I would like to thank my guide Mrs. M.Vidya,

Assistant professor, who guided me to complete this project
and also thank everyone those who supported me to achieve
my target.

REFERENCES
[1]
Fig.11Timing diagram
Area in (um2)

EXISTING

No. of
FFs
295

PROPOSED

76

0.912 um2

3.54 um2
[2]

Shiva Kumar Madishetty, Arjuna Madanayake,

Renato J. Cintra, VLSI Architectures for the 4-Tap
and 6-Tap 2-D Daubechies Wavelet Filters Using
Algebraic Integers, IEEE transactions on circuits
and systems, june 2013Feng M.L. and Tan Y.P.
(2004) Contrast adaptive binarization of low
quality document images IEICE Electron. Express,
vol. 1, no. 16, pp. 501506.
Michael Weeks and Qin Wang, Orthogonal
Wavelet Coefficient Precision and Fixed point
1134

IJRITCC | May 2014, Available @ http://www.ijritcc.org

_______________________________________________________________________________________

International Journal on Recent and Innovation Trends in Computing and Communication

Volume: 2 Issue: 5

ISSN: 2321-8169
1130 1135

_______________________________________________________________________________________________
[3]

[4]

[5]

[6]

[7]

representation Department of Computer Science,

Georgia University, 2001.
A. Ruiz, A. R. Arnauv and M. Ardinav, 2D
Wavelet Transform using Fixed Point Number
Representation, University of spain CICTY, 2001.
K. Wahid, V. Dimitrov, and G. Jullien, VLSI
architectures of Daubechies wavelets for algebraic
integers, J. Circuits Syst.Comput., vol.13, no.6,
pp. 1251-1270, 2004.
R. Baghaie and V. Dimitrov, Computing Haar
transform using algebraic integers, in Proc. Conf.
Signals, Systems Computers Record 34th Asilomar
Conf.,vol.1, pp. 438-442 ,2000.
B. K. Mohanty, A. Mahajan, and P. K. Meher,
Area and power efficient architecture for highthroughput implementation of lifting 2-ddwt,
IEEE Trans. Circuits. Syst. II, Jul. 2012.
K. A. Wahid, V. S. Dimitrov, G. A. Jullien, and W.
Badawy, An algebraic integer based encoding

[8]

[9]

[10]

scheme for implementing Daubechies discrete

wavelet transforms, in Proc. Asilomar Conf.
Signals, Syst. Comp., pp. 967-971, 2002.
K. A. Wahid, V. S. Dimitrov, and G. A. Jullien,
Error-free arithmetic for discrete wavelet
transforms using algebraic integers, in Proc. 16 th
IEEE Symp. Computer Arithmetic, pp. 238-244,
2003.
M. A. Islam and K. A. Wahid, Area- and powerefficient design of Daubechies wavelet transforms
using folded AIQ mapping, IEEE Trans. Circuits
Syst. II, pp. 716-720, Sep. 2010.
K. A. Wahid, V. S. Dimitrov, G. A. Jullien, and W.
Badawy, Error-free computation of Daubechies
wavelets for image compression applications,
Electron. Lett., pp. 428-429, 2003

1135
IJRITCC | May 2014, Available @ http://www.ijritcc.org

_______________________________________________________________________________________